The proposal made above introduces a mechanism for licensing resultative constructions and accounting for their semantics. As with any account of the resultative construction, this account must also include an explanation of the idiosyncrasy of the phenomenon. The representation utilised provides some of the necessary constraints, restricting path PPs to be a certain type and coercing locative PPs in a particular way. An additional constraint will be imposed by pragmatics, as will be discussed in Section 4.7.4. There will still, however, be much idiosyncrasy left unaccounted for, specifically cases in which syntactic and semantic substitutions lead to infelicity (see Section 4.3.3).
I mentioned in Section 4.3.3 that some mechanism for
encoding conventional constraints is needed in order to capture the
idiosyncrasy associated with resultative construction.
The framework provided by the Construction Grammar approach to this
phenomenon points to a solution. Constructions can be viewed as
specifying form-meaning pairs which are part of a language.
However, constructions are, like many other linguistic phenomena,
governed by conventional usage.
Goldberg (1995, p. 192) suggests there can be lexicalised instances of Constructions -- instances in which the form and the meaning specified in the Construction will be preserved, but which will define more precisely the particular words or class of words which can appear. These instances are analogous to idioms in that the structure, meaning and the lexical items which instantiate the construction are fixed, but they differ from idioms because they derive from a more generally available form-meaning pair.
The Resultative Construction can therefore be viewed as defining a semi-productive sense extension mechanism: verbs can be used in sentences with the form and meaning reflected in the Construction (as long as their semantics are compatible with the Construction and any other constraints are satisfied), but some of these sense extensions become lexicalised due to the conventions of language use in particular linguistic communities. In some cases these lexicalised instances are truly idiomatic. They may acquire a meaning which cannot entirely be predicted on the basis of the Construction: Does John laughed himself silly really mean that John became silly as a result of laughing? In You scared the daylights out of me, what are the daylights (Jackendoff )? Once there is a lexicalised instance of a construction involving a particular verb, it becomes difficult to use that verb in a different instance of the same construction because that use would conflict with the conventionalised form. Hence the oddness of examples like John laughed himself tired.
I suggest that the idiosyncrasy of the resultative construction is a reflection of the high degree of conventionalisation governing the construction. Speakers prefer instances of this construction which conform to their lexicalised instances. Many, if not most, occurrences of this construction which speakers use and come across reflect a lexicalised form. This fact accounts for the idiosyncrasy -- variations from lexicalised forms are in theory perfectly acceptable as they can be licensed by the existence of the Construction in the grammar but in practice they are viewed as anomalous or ungrammatical because they don't conform to the ``standard'' forms in use. Variations (i.e. syntactic and semantic substitutions) of entirely novel instances of the construction are therefore tolerated much more easily than variations of highly colloquial instances. So I can easily accept all variants in resalt32, but the variants of the colloquial resalt33a-b in resalt33c-d are less felicitous.
Sue brushed her hair smooth. Sue brushed her hair shiny/straight/flat/out of her eyes/...
Sue cried herself to sleep. Sue cried herself sick. ?Sue cried herself asleep. *Sue cried herself sleepy.
This conventionalisation is itself a result of the fact that the resultative constructions are licensed by a form-meaning pair in the grammar. These constructions are very different from constructions which are interpreted strictly compositionally, in that they don't result from generative mechanisms in the grammar and in that their meaning is essentially fixed. These properties indicate that Constructions are ``special'' in grammatical terms -- that is, they do not follow from the normal principles of grammar. That they are subject to a much higher degree of conventionalisation than other constructions seems natural given that their existence in the grammar can be viewed as a result of conventionalisation.
This perspective on the Resultative Construction is in line with observations made about the semi-productivity of many other lexical processes e.g. the generation of denominal verbs (Jackendoff ), and blocking by exceptional forms in sense extensions (Briscoe et al. briscoe_etal:95, Copestake and Briscoe 1995). The property of semi-productivity has been argued to require lexicalisation of the forms output by the lexical rule. Furthermore, preemption by synonymy (Copestake copestake:95), in which an extended meaning will not be conventionalised if a common synonym exists, has been shown to be overridable in context in that a blocked form can be interpreted. Both of these characteristics surface in resultative constructions, indicating that it is a generative process constrained by conventionality.
In sum, I assume that there are lexicalised instances of this Construction represented in the lexicon, and that these instances limit the acceptability of sentences which vary from the conventionalised pattern.